Liver tumor segmentation from computed tomography images using multiscale residual dilated encoder‐decoder network

In this paper, an encoder‐decoder‐based architecture, which segments liver tumors with a two‐step training process is proposed. Accurate liver tumor segmentation from CT images is still a major problem that impacts the diagnosis process. Heterogeneous densities, shapes, and unclear boundaries make t...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:International journal of imaging systems and technology Jg. 32; H. 2; S. 600 - 613
Hauptverfasser: Tummala, Bindu Madhavi, Barpanda, Soubhagya Sankar
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Hoboken, USA John Wiley & Sons, Inc 01.03.2022
Wiley Subscription Services, Inc
Schlagworte:
Coders
Computed tomography
Data loss
Datasets
dilated convolutions
encoder‐decoder architecture
Image segmentation
Liver
liver tumor segmentation
Medical imaging
semantic segmentation
Training
Tumors
ISSN:0899-9457, 1098-1098
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In this paper, an encoder‐decoder‐based architecture, which segments liver tumors with a two‐step training process is proposed. Accurate liver tumor segmentation from CT images is still a major problem that impacts the diagnosis process. Heterogeneous densities, shapes, and unclear boundaries make tumor extraction challenging. First, the proposed network segments the liver, and then tumors are extracted from the liver ROIs. We have scaled down the images into different resolutions at each scale and applied normal convolutions along with the dilations and residual connections to capture broad conceptual information without data loss. MDICE, a combined loss function is used to enhance the learning capability and the 3D‐IRCADb1 dataset is considered for training and testing because of its tumor complexities. The segmentation quality metrics DICE, MDICE are analyzed on the 3D‐IRCADb1 dataset and obtained 0.98 and 0.65 accuracies per case for liver and tumor segmentation respectively, and found improvement over U‐Net and other variants.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0899-9457
1098-1098
DOI:10.1002/ima.22640